Community genetics in the Northwestern Atlantic intertidal

Authors


J. P. Wares. Present address: Center For Population Biology, University of California of Davis. Fax: +1-505-277-0304; E-mail: jpwares@ucdavis.edu

Abstract

Our ability to make inferences about the processes acting upon a biological system can be dramatically improved through integration of information from other fields. In particular, this is true of the field of phylogeography, a discipline that attempts to describe geographical variation in species using neutral genetic diversity as a correlate of time. Through comparisons of information from multiple species, as well as background information about the abiotic environment and how it has changed over time, we should be able to reassemble many of the forces that were important in assembling the communities and community interactions found in a given region. Here I review the information available for coastal species in the northwestern Atlantic, and argue that an integration of ecological, genetic, geological and oceanographic information can illustrate emergent patterns of community genetics.

Introduction

It is becoming clear that integration of knowledge across scientific fields is the new standard for increasing our awareness of the processes shaping our natural world (Wilson 1998, Elser et al. 2000). For example, our understanding of community dynamics is shaped by our observations of the behavioral and trophic interactions within and among species that share an ecosystem; these interactions are not only complex, they are dynamic and evolving. Integrating information about these interactions, the extrinsic forces acting on members of the community, and the genealogies describing historical and contemporary migrant exchange and demographics within species will direct our understanding of biotic responses to environmental perturbation (National Science Foundation 1998, Agrawal 2001).

Such integration has been initiated in the field of phylogeography (see Avise 2000), which attempts to describe geographic variation in species using neutral genetic diversity as a correlate of time. The true promise of this discipline, however, lies in the ability to compare the evolutionary and ecological backgrounds of a number of species (Cunningham and Collins 1998, Avise 2000, Wares and Cunningham 2001). These comparisons can indicate how communities are assembled, and how ecological interactions affect the association of species in time and space (Bertness and Callaway 1994). While the true phylogeographic history of a single species can be elusive without extensive sampling of individuals and independent genetic loci (Wollenberg and Avise 1998), it is clear that only a coordinated search across a number of interacting taxa can clarify the life history traits that partially determine species responses to environmental change and the presence of other taxa (e.g.Wares and Cunningham 2001).

Here I argue that although it is rare to find a number of regional phylogeographic studies that agree on what mechanisms are involved, deterministic processes — such as the differing responses of species to climatic change due to variation in life history traits — do exist in natural communities. This review will focus on the Altantic coast of North America as a region in which an integration of ecological, genetic, geological, and oceanographic information can illustrate emergent patterns of community genetics. I will first review what is known about the geology, oceanography, and historical environment for this region, which suggests that the Northwest Atlantic Boreal Shelf is a useful empirical system for integrated research. Then I ask how these factors are related to phylogeographic issues and review the work that has been done in this region in the context of its biotic and abiotic history.

The Northwest Atlantic boreal shelf

The Atlantic coast of North America is of particular interest because a number of geologically recent and well-characterized historical events have moulded the biota of this region (Flint 1940). Many of the species found in this region arrived only in the late Pliocene (Vermeij 1991), and since then the combination of changing habitat, Pleistocene glaciation and oceanographic forces has had a strong impact upon the distribution and abundance of these species (Pielou 1991). The emphasis of this study is on the temperate boreal shelf region (Briggs 1974), which includes the Virginian biogeographical province and the overlap near Cape Cod with the Nova Scotian province (Hayden & Dolan 1976). Within this region are the Lower and Upper Virginian subprovinces (Engle & Summers 1999), distinguishable by temperature, salinity and faunal composition.

Although Cape Hatteras is clearly an area of strong biogeographical transition in the western Atlantic (Hayden & Dolan 1976; Gaylord & Gaines 2000), the near-shore intertidal and estuarine patterns of transition within and among species are less clear (Engle & Summers 1999; Jones & Quattro 1999). This discussion instead features biotic distributions to the north that may have formed because of both the historical effects of Pleistocene glaciation and strong environmental gradients. The only widely recognized geographical feature associated with faunal transition in this region is Cape Cod.

Like many other provincial boundaries, the distribution of species range limits at Cape Cod is asymmetric, with more species reaching their northern limit at Cape Cod than species reaching their southern limit (Engle & Summers 1999). A variety of explanations have been proposed for the distribution of species and intraspecific genetic lineages around other asymmetric transition zones.

Hypothesis 1: Habitat distribution

Common oceanographic processes may lead to a prevalence of one particular habitat type (e.g. cooler water) on both sides of a biogeographical transition zone for short distances; although at larger geographical scales, the habitat type is more distinct on each side. Owing to a variety of mechanisms including meso-scale eddies and up-welling, water/mass boundaries are often more effective northern barriers to species distributions than they are barriers to southward range expansion (Roy et al. 1995). Water temperature increases to the south of Cape Cod, and may represent a primary mechanism for differentiation of species groups found in the Virginian and Nova Scotian provinces. However, pockets of cooler water are found to the south of Cape Cod, and the cold temperate waters to the north are generally much more stable in temperature than the warmer waters to the south (Engle & Summers 1999), a distribution that lends support to this hypothesis. The exceptions to this pattern lie in the waters at the head of the Bay of Fundy and in the Gulf of St. Lawrence, warmed by high levels of insolation (Bousfield & Thomas 1975).

Hypothesis 2: Ocean currents

Ocean currents may be responsible for maintaining asymmetric biogeographical transition zones. Currents are partly responsible for maintaining thermal patterns, and may influence the distribution of species themselves. The predominant currents in this region sweep south−south-west across Cape Cod and Long Island Sound (Bumpus 1973; Mayer et al. 1979), suggesting that currents could limit the northward spread of some species or intraspecific variants and that passive larval dispersal acts to maintain populations of cold temperate species south of Cape Cod. Intraspecific genetic patterns can respond to local ocean currents and eddies (Bucklin 1991; Rocha-Olivares & Vetter 1999; Wares et al. 2001), although it is often a subtle response. Local currents include the Virginia coastal current, originating in Long Island Sound with a predominantly southward flow, in the same general direction as the offshore Labrador current (Bousfield & Thomas 1975; Cronin 1988). Tidal currents in this transitional region are also predictable (Moody et al. 1984), but the effect of these oceanographic forces on dispersal depend on the specific spawning and settlement cues used by each species (Gaylord & Gaines 2000).

Hypothesis 3: Glaciation

The most effective force that has shaped the historical distribution of species in this region is Pleistocene glaciation. Glacial maxima covered much of the New England coast with sheets of ice over 1 km thick (Pratt & Schlee 1969; Shackleton et al. 1984; Pielou 1991). Although this activity rendered much of the coastline uninhabitable, concomitant changes in water temperature pushed the geographical distribution of many species to the south (Cronin 1988; Bucklin & Wiebe 1998). Therefore, we would expect that not only would species diversity be higher to the south of the glacial margin, but also genetic diversity will be higher in southern populations of species that have recolonized the glaciated region (Hewitt 1996, 2000). Whereas some species may have been extirpated from the region, others survived in unglaciated refugia both to the south of the glacial margin and to the north, in Newfoundland and Nova Scotia (Dyke & Prest 1987; Holder et al. 1999; Fig. 1). A remaining challenge is the reconstruction of available habitat at different times relative to glacial maxima (Pielou 1991); for instance, lower sea levels (≈ 150 m) probably exposed consolidated sediments on the continental shelf (Riggs et al. 1996), which could act as habitat for rocky intertidal species.

Figure 1.

The region along the Atlantic coast of North America between Delaware and Massachusetts may represent a significant zone of phylogeographical transition. A sliding window analysis was performed to see if available phylogeographical data are consistent with a single focal region of phylogeographical transition. The distribution of each genetic type (described by the original authors of each study as a distinct clade or allele frequency pattern) was compiled, with the assumption that the phylogeographical discontinuity existed somewhere between the northern-most population of the ‘southern’ type and the southern-most population of the ‘northern’ type. This allows for the two distributions to overlap. These data were sampled in a series of window increments, in which a coastal region of size x (km) was analysed for the number of possible phylogeographical discontinuities that could lie within that region. The window then moved stepwise along the coast, each time both calculating the number of possible discontinuities given the actual data, and performing a series of 1000 randomizations of the geographical limits of both southern and northern ‘types’ to produce a series of random phylogeographical discontinuities. The proportion of Monte Carlo replicates with a larger number of possible discontinuities than the actual data was calculated, and geographical regions where this proportion P < 0.05 are shaded in grey (where x corresponds to ≈ 10 min latitude; the size of the indicated region was relatively insensitive to x). This is not a definitive description of where the phylogeographical discontinuity lies in each data set, but indicates a region of high potential in terms of describing such genetic patterns in species of the Atlantic coast. The species included are the isopod Idotea balthica* (a), the hermit crab Pagurus longicarpus* (b), the ocean quahog clam Arctica islandica (c), a red alga Chondrus crispus (d), an amphipod Ampithoe longimana* (e), the coastal tiger beetle Cicindella dorsalis* (f), the killifish Fundulus heteroclitus* (g), polychaetes Marenzellaria viridis* (h), seastars in the genus Asterias* (i), the marsh grass Spartina alterniflora* (j), the isopod Cyathura polita (k), fishes Acipenser oxyrhinchus, Morone saxatilis (l, m), the clam Mercenaria mercenaria (n), copepods Eurytemora affinis* (o), the barnacle Semibalanus balanoides (p), the American Shad Alosa sapidissima (q) and slipper limpet Crepidula convexa (r). References are in Table 1, and those species marked by an asterisk were used for the Monte Carlo analysis, whereas others are included as illustration of the consistency across a number of taxa in this region. Bars only represent the sampled distribution of genetic ‘types’, not the species distribution. Maps adapted from Dyke & Prest (1986).

Table 1.  Atlantic coastal species that have been analysed for population genetic divergence that are distributed in the study region. Limits are based on sampling, not on actual species/type distribution, and groups are as delimited by authors or as inferred from cladistic structure in each study; some groups are only minimally divergent (e.g. Mercenaria populations only differ at one of seven allozyme loci) but for illustrative purposes all are included. Only species with distinct phylogeographical clades, based on DNA sequence data, and exact sampling locations were used for the Monte Carlo analysis illustrated in Fig. 1. IBD indicates populations that have been tested for isolation by distance or that are quite likely to exhibit such a pattern if tested; ‘NS isolate’ indicates those species with populations in Nova Scotia that are genetically divergent
SpeciesClassificationN limit of S groupS limit of N groupNS isolate?Reference
Idotea balthicaArthropoda: Isopoda37°14′41°26′YesWares (2001a)
Pagurus longicarpusArthropoda: Decapoda34°43′41°31′YesYoung et al. in press
Arctica islandicaMollusca: Bivalvian/an/aYesDahlgren et al. ( 2000)
Chondrus crispusRhodophyta: Gigartinalesn/an/aYesChopin et al. (1996)
Ampithoe longimanaArthropoda: Amphipoda38°46′38°46′Sotka et al. in review
Cicindella dorsalis dorsalisArthropoda: Hexapoda38°29′41°25′Vogler & DeSalle (1993)
Fundulus heteroclitusTeleosti: Cyprinodontiformes40°30′39°22′Smith et al. (1998)
Marenzellaria viridisAnnelida: Polychaeta43°07′38°47′NoBastrop et al. (1998)
Asterias rubens/forbesiEchinodermata: Asteroidea41°3135°22′NoFranz et al. (1981), Wares (2001b)
Spartina alternifloraAnthophyta: PoaceaeIBDIBDO’Brien & Freshwater (1999)
Cyathura politaArthropoda: Isopoda36°11′41°23′Brown et al. (1988)
Acipenser oxyrhinchusTeleosti: Acipenseriformes32°02′40°41′Waldmann et al. (1996)
Morone saxatilisTeleosti: Perciformes39°25′40°41′Waldmann et al. (1996)
Mercenaria mercenariaMollusca: Bivalvia38°30′41°25′Dillon & Manzi (1987)
Eurytemora affinisArthropoda: Copepoda38°59′41°23′NoLee (1999)
Semibalanus balanoidesArthropoda: Cirripedian/an/aNoWares & Cunningham (2001), Brown et al. (2001)
Alosa sapidissima1996Teleosti: ClupeiformesIBD?IBD?Bentzen et al. (1989), Waldmann et al. (1996)
Crepidula convexa (Northern sp.)Mollusca: Gastropoda31°04′38°56′Collin (2001)
Tautoga onitisTeleosti: Perciformesn/an/aOrbacz & Gaffney (2000)

Phylogeographical methods have promise for separating the roles of different historical forces in shaping a community, because genetic relationships within a species are likely to reflect the action of these forces, particularly when a species’ geographical range includes a significant biogeographical boundary (Avise 1992; Wares et al. 2001). By comparing a broad range of marine organisms in the boreal shelf region, we can access both the particular life history traits that allow species to exploit a changing distribution of suitable habitat, and the genetic data that describe the recent history of these species. These comparisons allow more detailed inferences about the historical intertidal community in this region and how it responded to Pleistocene climate change.

Are phylogeographical and biogeographical transitions comparable?

Biogeographical patterns are made clear by the concordant geographical distributions of many species (Briggs 1974; Nelson & Platnick 1981). By the same reasoning, finding concordant phylogeographical patterns, whether due to vicariance, range expansion or similar processes, strengthens the inference for a mechanism that broadly acts on species dynamics. Comparisons of broad biogeographical patterns with phylogeographical data may be used to infer the relative strength of historical events and ongoing responses to the environment in shaping regional diversity (Burton 1998). Broad biogeographical patterns have been described for this region that involve the late Pliocene arrival of the Northwest Atlantic biota and the distribution of distinct groups of species. The relationship between these faunal patterns and the environment may be determined by ecological requirements of individual species or by underlying historical patterns of species isolation and species coevolution (Endler 1982). We do not know, in most instances, if we are observing a process of ecological determination (in which intrinsic traits of a species are driving the divergence), or if environmental factors are simply maintaining a historical pattern of species distributions that were established long ago by different causes. Intraspecific genetic discontinuities (‘phylogeographical breaks’ or ‘genetic subdivision’), such as species range boundaries, are drawn toward environmental discontinuities (Endler 1977; Gaylord & Gaines 2000). Partial barriers to gene flow can cause genetic clines (or ‘tension zones’, Barton & Hewitt 1989) to be attracted to them, even when physiological optima are quite different across species or genetic loci (Endler 1977).

Thus, the two fields are quite similar — but the temporal element of phylogeography allows us to better discriminate among potential mechanisms. If the patterns of biogeographical transition and phylogeographical discontinuity coincide, we can trust that ongoing or recent processes are in part responsible for driving patterns of species distribution and abundance (Meyer & Paulay, manuscript submitted). When they do not coincide, the mechanisms that may be involved in generating a phylogeographical pattern probably have little to do with the biogeography of the region and may reflect the idiosyncratic history of a particular species. In addition to spatial coincidence, the depth of the phylogeographical break is informative. Comparisons of genetic patterns in four species that span the biogeographical transition zone at Point Conception, California (Wares et al. 2001) showed that regional ocean currents could explain the co-distribution of so many species range boundaries, but not because these patterns indicated a concordant phylogeographical ‘break’ (Burton 1998). Instead, a cladistic analysis of migration among populations on either side of the transition zone indicated that gene flow appears to be driven by strong offshore currents. In cases where there is a deeper genetic discontinuity among populations of a single species, additional factors must be considered. In these cases, the distinction must be made between populations that are locally adapted to the distinct environments on either side of the boundary, effectively limiting gene flow, and those deep divergences that represent historical signals that gene flow has not yet eliminated. Examples of each are discussed below (see Asterias and Ampithoe). This distinction is fundamental to phylogeographical hypothesis testing; it involves discrimination between populations that have reached an ecological and genetic equilibrium (i.e. genetic diversity proportional to population size) in their environment, and those that have not.

Species interactions may also drive the co-distribution of biogeographical and phylogeographical discontinuities. Predatory or competitive interactions may limit a species’ geographical range (Case & Taper 2000), whereas interactions that ameliorate a harsh environment may positively influence a species’ distribution (Bertness et al. 1999). Increasing the coordination between phylogeographers and ecologists is just one direction that must be taken to improve our understanding of community responses to environmental change. How often are interactions with other taxa as fundamental to the distribution and abundance of an intertidal species as historical events and environmental gradients? In what complex ways do abiotic and biotic influences interact to generate the patterns of diversity in present regional assemblages? These interactions can, at a basic level, be described by parameters including historical and current Ne and m, the genetic effective population size and population migration rate, suggesting that phylogeographers must continue to closely consider the meaning of the phrase ‘molecular ecology’.

Phylogeographical concordance in the Northwestern Atlantic

Whereas true meta-analysis of most published population genetic data is difficult (Leberg & Neigel 1999), a review of the work that has been carried out in this region shows that there are similar genetic patterns across a number of species. A number of studies have characterized genetic discontinuities in Atlantic coastal species, and the geographical location and scale of these phylogeographical breaks can be summarized. Figure 1 illustrates the distribution of intraspecific genetic variation that has been found in studies of allozyme and sequence divergence over the past 30 years. Many of these studies, which include algal, invertebrate and vertebrate species, are represented by deep phylogeographical breaks (type I pattern, Avise 2000) or strong allelic frequency clines. Monte Carlo resampling (see legend, Fig. 1) of these genetic−geographical relationships indicates that there are more genetic discontinuities between 39 and 42° N than would be expected at random, even considering the patchy geographical sampling in some of these data sets (including Wares 2001a, 2001b).

This significant transition roughly corresponds to the major provincial boundary between the Nova Scotian and Virginian provinces (Hayden & Dolan 1976), but many of these phylogeographical transitions are actually displaced to the south of Cape Cod. Because this association may be more closely linked to the boundary between the Upper and Lower Virginian provinces (Engle & Summers 1999), we can ask what factors appear to be most important in determining such a distribution. This minor provincial boundary, at ≈ 39° N latitude (bisecting Delaware), marks a strong gradient in average water temperature, and the benthic invertebrate faunal composition in the Upper Virginian is more similar to the fauna of Long Island Sound and the Nova Scotian province than to the species found immediately to the south (Engle & Summers 1999). It is clear that genetic discontinuities in several species (including the amphipod Ampithoe longimana, the killifish Fundulus heteroclitus, the polychaete Marenzellaria viridis and the copepod Eurytemora affinis) are more closely affiliated with this region than with Cape Cod to the north.

The relationship of thermal boundaries and phylogeographical discontinuities with respect to the biogeographical boundary at Cape Cod supports the water/mass boundary hypothesis of Roy et al. (1995; see Hypothesis 1 above). Meso-scale oceanographic forces may tend to distribute cooler waters from the Nova Scotian province to the south, and there may be few opportunities for species or genetic lineages from the Lower Virginian to migrate northward. But is this because of actual limits to migration, suggesting that ocean currents are responsible? As noted above, there are some warm-water populations of temperate species in the Nova Scotian province (Bousfield & Thomas 1975; Clark & Downey 1992; Dillon & Manzi 1992). There is also a historical pattern of shifting isotherms to be considered; the boundary between habitats of cold-temperate and mild-temperate species is now located around 42° N latitude, but this was as far south as 37–37.5° N during the late Pleistocene (Cronin 1988). A closer look at the phylogeographical history of the species in this review may help distinguish between patterns formed by historical fluctuations in the distribution of suitable habitat, those patterns that have been formed by adaptation to the present environment and those that are being shaped by coastal currents that may push some species boundaries and genetic clines to the south of Cape Cod, where they coincide with the thermal and salinity gradients of the Upper/Lower Virginian boundary. Because these three mechanisms represent distinct balances among equilibrium populations with populations that reflect a genetic disequilibrium due to historical changes in distribution, I briefly review how these types of patterns can be recognized.

Equilibrium and nonequilibrium genetic patterns

Marine populations are constantly changing in size and geographical distribution because of climatic and environmental shifts. The effect of these changes on the genetic effective population size (Ne) and rates of migration from one region to another is essentially to prevent most marine species from attaining true equilibrium between gene flow and genetic drift (see Avise 2000; Grosberg & Cunningham 2001 for thorough reviews of these topics). Genetic expectations given a dramatic disequilibrium include lower genetic diversity than expected given the population census size or levels of migration that are far greater or far less than expected given Ne. These results might be found for populations that have been recently formed because of geographical range expansion, or when external forces (such as ocean currents) are affecting migration patterns. In each case, the genetic differentiation between populations will be very subtle (e.g. Bernatchez & Wilson 1998; Wares et al. 2001).

Shallow population genetic structure (slight genetic differentiation among populations) may also represent a near-equilibrium history. When per-generation dispersal capabilities are low, populations may attain a strong relationship between genetic divergence and the geographical distance between them (isolation by distance; Neigel et al. 1991; Slatkin 1993; Hellberg 1994). Isolation by distance generally suggests that populations are more nearly at equilibrium with respect to gene flow and genetic drift (Hutchison & Templeton 1999; Grosberg & Cunningham 2001; but see Pogson et al. 1995), and suggests that, other than the species’ dispersal capacity, there have been no physical or environmental barriers to gene flow for a substantial period.

Because isolation by distance is often a null hypothesis of phylogeographical equilibrium, and is easy to assay given appropriate geographical sampling and sample sizes (Hellberg 1994; Hutchison & Templeton 1999), this should be one of the first steps in any phylogeographical analysis. Because genetic distances among populations tend to increase with geographical distance, patchy geographical sampling may lead to spurious inferences of deeper phylogeographical breaks. In the marsh grass Spartina alterniflora, O’Brien & Freshwater (1999) intentionally sampled only a few regions along the Atlantic coast to examine the hypothesis, based on variation in phenology among regions, that populations just south of Long Island Sound would be more similar to Carolinian populations than to New England populations. Appropriate analysis of the data shows a consistent pattern of isolation by distance within and among regional samples (O’Brien & Freshwater 1999), yet the gene tree produced from these data otherwise appears consistent with deep genetic discontinuities among these three regions. Such deep divergences would require a significant period of reproductive or geographical isolation to become established (see Information box). Whether population isolation is due to historical mechanisms (e.g. vicariance) or adaptation and reproductive isolation, less genetic diversity is lost in these primary and secondary divergence events than in ‘nonequilibrium’ cases such as founder events associated with range expansion. Because populations separated by deep phylogeographical ‘breaks’ are older, they have not gone locally extinct, hence the deep stable divergence, they are more nearly at genetic equilibrium than recently founded populations.

Many of the deep divergences observed on the Atlantic coast of North America, however, are due to both nonequilibrium processes and historical isolation. Many amphi-Atlantic species have recently colonized North America from European populations, often as recently as the Holocene epoch (Wares & Cunningham 2001). For these species, patterns consistent with geographical range expansion must be considered simultaneously with patterns of population isolation in North America. It is often difficult, however, to unambiguously determine the temporal patterns of range expansion and secondary contact among populations that have very recently changed in size or distribution. Some studies (e.g. Templeton et al. 1995; Austerlitz et al. 1997) suggest that the distribution of coalescent times among lineages should help detect former colonization events, but there is typically little allelic diversity associated with these expansion populations because so little time has passed for unique mutations to have arisen (Maruyama & Fuerst 1984; Templeton 1998; Wares & Cunningham 2001).

The establishment of strong population isolation, by any mechanism, will not necessarily produce a geographical-genetic pattern that is concordant with ecological and oceanographic patterns. Working against large-scale historical events that have occurred in this region are large-scale disturbances (e.g. hurricanes) which may dramatically affect larval distribution patterns from one year to the next. Under conditions of gene flow and drift alone, alterations in the pattern of larval recruitment for even a few generations may determine the general spatial pattern of gene frequencies for the next few hundred generations (Endler 1977; Bucklin & Wiebe 1998).

History, adaptation and gene flow

Despite the potential for other forces, including adaptation or ocean currents, to form the general patterns of phylogeographical discontinuity seen in marine species of the boreal shelf, the initial process of isolation among these populations is likely due to Pleistocene glaciation. Since the last glacial maximum ≈ 20 000 years ago, there has been little interruption in the mud flat–estuarine habitat that is distributed almost continuously from Long Island Sound to the North Carolina coast (Emery & Garrison 1967; Blackwelder 1980; Fletcher et al. 1990; Engle & Summers 1999). Just prior to the Wisconsin maximum (30 000–35 000 years ago), sea levels and interglacial temperatures were similar to present conditions (Milliman & Emery 1968). Lower sea levels during glacial maxima probably did not persist long enough to permit significant population differentiation (but see Palumbi 1994 for a discussion of gametic mechanisms that promote rapid divergence). Deep genetic divergences among populations were more likely initiated by older, more geographically widespread events, including the extirpation of populations between glacial refugia in North America as well as Europe (see Information box).

In addition to refugial populations of many Atlantic taxa existing south of the glacial margin in both North America and Europe, extensive terrestrial and coastal regions in the Canadian Maritimes and Iceland remained unglaciated (Berggren & Hollister 1974; Denton & Hughes 1981; Dyke & Prest 1986, 1987; Piper et al. 1990; Bernatchez & Dodson 1991; Holder et al. 1999). As with other benchmark studies of comparative phylogeography (e.g. Avise 1992), intraspecific genetic patterns have been used to confirm the geological predictions of the location and age of these refugia (Holder et al. 1999). At the western end of Nova Scotia, the faunal diversity is significantly higher than in surrounding regions (Bousfield & Thomas 1975), suggesting that a number of species that were extirpated from other regions never disappeared from this area; the studies reviewed here support this as well.

Many amphi-Atlantic species exhibit much higher genetic diversity in European populations than in North America (Wares & Cunningham 2001), a signal that these American populations probably represent postglacial expansion events. Nevertheless, some isolates exist that apparently survived in the periglacial refugia; these populations may also have significantly reduced genetic diversity but should be more genetically diverged from ‘source’ populations in Europe or elsewhere (see Information box). The isopod Idotea balthica, in addition to evidence for expansion from Europe, is genetically diverse throughout its North Atlantic range (Wares 2001a). Populations in Virginia, Nova Scotia and Iceland are all significantly divergent from the primary amphi-Atlantic lineage. Each of these populations likely represents a distinct refugium. Other species share this refuge-associated phylogenetic diversity: the hermit crab Pagurus longicarpus (Young et al. manuscript submitted), the quahog Arctica islandica (Dahlgren et al. 2000) and the red alga Chondrus crispus (Chopin et al. 1996) all contain divergent Nova Scotian lineages (Fig. 1). The mussel Mytilus also fits this pattern (C. W. Cunningham, personal communication), and the putative Icelandic refugium may be evidenced by close genetic ties between Icelandic and North American populations of Mytilus and the barnacle Semibalanus balanoides (Wares & Cunningham 2001 appendix B; C. Henzler, personal communication).

The sharp genetic breaks seen in American endemic taxa are probably related to the same historical isolation events. However, we must balance this view with the possibility that adaptation to distinct environments (e.g. water temperature or salinity) has either formed or is maintaining the observed divergence (Hypothesis 1, habitat distribution). The killifish Fundulus heteroclitus has been thoroughly studied for ecological and phylogeographical variation (reviewed in Smith et al. 1998) and exemplifies this balance. Using dense population sampling and a number of genetic loci, research has shown distinct northern and southern genetic types. The boundary between these types localizes to the New Jersey coastline (Smith et al. 1998), but extant populations of the northern group can be found in the northern reaches of Chesapeake Bay and the Delaware Bay estuaries. These populations are apparently relicts of a time when the boundary between the two types was shifted southward by environmental temperature change, and the northern type was able to disperse into these regions—indicating that whether this boundary has formed by primary divergence (selection) or secondary contact following vicariance, the formation of this boundary predates the most recent glacial maximum.

Other species also respond to variation in thermal habitat, but not always with concordant phylogeographical histories among species. Two clades of the polychaete Marenzellaria viridis are distributed along the American coast and differ in habitat salinity with the separation between these clades, however, mostly concordant with the Upper/Lower Virginian break (Bastrop et al. 1998), blurring the line between history and ecology as determining factors in this phylogenetic history (Endler 1982). The species pair Asterias rubens and A. forbesi are also divergent genetically and in thermal tolerance (Franz et al. 1981). In this case, physiological adaptation is not driving divergence, but rather maintaining it, as the historical divergence between these species is unrelated to the current biogeographical transition (Wares 2001b). Mytilus edulis responds to salinity gradients at allozyme loci (reviewed in Hilbish 1996) with no correlation to neutral mitochondrial diversity, however, and a similar lack of concordance is observed in the barnacle Semibalanus balanoides. The allozyme locus Mpi has been shown to vary among comparisons of microhabitats with high vs. low thermal stress (Schmidt & Rand 1999), but neutral mitochondrial markers do not co-segregate in these comparisons. There are apparently two distinct genetic lineages of Semibalanus in this region (Wares & Cunningham 2001), and high dispersal (via planktonic larvae) from one population to the next has probably contributed to the rapid homogenization of the frequency of each type throughout the North American range of this barnacle (Brown et al. 2001).

Of the species with lower dispersal potential, some can be readily associated with historical separation in glacial refugia, whereas others exhibit genetic patterns more consistent with equilibrium isolation by distance. Examples of the first group include the amphipod Ampithoe longimana and the coastal tiger beetle Cicindella dorsalis. Each exhibits distinct genetic subdivision between populations north of and including Long Island Sound, and those populations to the south. Sotka et al. (in review) compared putatively neutral genetic markers with adaptive genetic traits in Ampithoe. In this species, diet preferences in populations that are sympatric with the brown alga Dictyota are adaptive for predator avoidance, but disappear in northern populations of Ampithoe, where Dictyota is not found. The deep phylogeographical break in Ampithoe, concordant with the Upper−Lower Virginian transition zone, does not match the clinal pattern of behavioural adaptation to Dictyota. This suggests that moderate gene flow throughout the range of Ampithoe is not high enough to overcome a combination of historical divergence between the northern and southern populations of Ampithoe and an apparent selection gradient based on the maintenance and costs of Dictyota tolerance (Sotka et al. in review).

The dispersal ability of a species has a direct but often complicated effect on the phylogeographical relationships within that species (Grosberg & Cunningham 2001; Wares & Cunningham 2001). Similarly, there are relationships between the type of species that are found as specialists in rocky intertidal habitats and those that are habitat generalists; rocky intertidal specialists are typically amphi-Atlantic taxa (Bousfield & Thomas 1975; Ingólfsson 1992), suggesting that different forces might generate phylogeographical patterns in these species than in habitat generalists, or in American endemics in general. The interaction of characteristics such as these, dispersal ability, habitat preferences and species interactions, are likely to play a deterministic role in phylogeography, although they have not been adequately studied in a comparative programme.

Reaching the limits of the data that have been collected, we are now left with those species that exhibit more shallow population structure or inadequate data for phylogeographical hypothesis testing. This group of studies may be considerably harder to characterize because of a certain level of publication bias; invariant data may not be reported in the same way as data that indicate deep genetic subdivision. Nevertheless, low genetic variation within and among populations may be indicative of either important demographic processes (Grant & Bowen 1998) or historical events (Orbacz & Gaffney 2000). Some data sets may also be inadequate for distinguishing between historical isolation and equilibrium isolation by distance (Templeton et al. 1995). O’Brien & Freshwater (1999) illustrate this in Spartina, as mentioned above. Reproductive traits in this species are likely to isolate distant populations from each other, not only in terms of gametic dispersal, but also phenological differences among populations. Even given high dispersal capabilities of gametes, larvae or juveniles, these differences may not isolate nearby populations reproductively, but will limit regional gene flow. Species that use environmental cues for spawning may tend toward limited gene flow, and this characteristic should be considered for integrative community genetics.

A number of other studies suggest that significant divergence among populations localizes to the region highlighted in Fig. 1, but many of these studies reported that these data answer different, nonphylogeographical questions, and thus isolation by distance has not been explicitly analysed for some of the species. These species include isopods (Cyathura polita, Brown et al. 1988), fishes (Acipenser oxyrinchus, Alosa sapidissima and Morone saxatilis, Bentzen et al. 1989; Waldman et al. 1996), the clam Mercenaria mercenaria (Dillon & Manzi 1992), and the horseshoe crab Limulus polyphemus (Pierce et al. 2000). Within the Atlantic clade of the coastal copepod Eurytemora affinis, basal alleles are all found south of Chesapeake Bay, with a well-supported clade nested in this group that consists of individuals from Massachusetts and Canada (Lee 1999). Similar geographical sampling has produced similar patterns in the isopod Idotea balthica (Wares 2001a) and the ‘northern species’ of the slipper limpet Crepidula convexa (Collin 2001), although with deeper genetic divergence. Improved geographical sampling of these and many species, accompanied by more intense ecological scrutiny of these populations, will help determine the validity of the phylogeographical transition zone illustrated in this paper. Common signals seen in these data sets, however, include decreases in allelic diversity in northern populations (Mercenaria mercenaria, Dillon & Manzi 1992; Alosa, Bentzen et al. 1989; Nolan et al. 1991; Waldman et al. 1996; Acipenser, Waldman et al. 1996; Fundulus, Smith et al. 1998; also see Bernatchez & Wilson 1998). This is a commonly cited effect of postglacial range expansion (Hewitt 1996, 2000), and the inferences of range expansion that can be made (Templeton et al. 1995; Austerlitz et al. 1997; Johnson et al. 2000) are critical for distinguishing among different types of isolation in the Northwest Atlantic (Hypothesis 3, glaciation).

The highlighted region in Fig. 1 may indicate that processes of history and adaptation have played a significant role in generating phylogeographically distinct clades or populations within a number of species along the Atlantic coast. Geographic variation in ecological interactions may also determine the overall distribution of species or ecotypes along this coast (Woodin 1983; Wilson 1991; Bertness & Callaway 1994; Travis 1996; Agrawal 2001) or even drive diversification (Doebeli & Dieckmann 2000), and it remains for the interactions among researchers to resolve this.

Conclusions

This review is not intended as a criticism of previous population genetic work from the Atlantic boreal shelf region, but as a stimulus for further study. Here, I outline a single region in which a diverse group of species shares a generalized pattern of phylogeographical transition. With improved sampling of a diversity of taxa and the application of recently developed methods for dissecting phylogeographical history in greater detail (e.g. Kuhner et al. 1998; Nielsen & Wakeley 2001), the link between genetic patterns and the ecology and distribution of a species relative to habitat distribution will become more apparent.

Unfortunately, this brief review is not able to resolve even the simple trichotomy of possible explanations that have been offered for generating the phylogeographical pattern common to these species. Testing hypotheses of oceanographically mediated dispersal (Hypothesis 2, ocean currents) will require the use of methods that can estimate asymmetric gene flow patterns (e.g. Beerli & Felsenstein 1999; Wares et al. 2001), and closer links between the study of adaptive traits and neutral genetic diversity (e.g. Sotka et al. in review) will generate more information about the role of selection in generating or maintaining genetically divergent populations.

Nevertheless, the historical influence of geographical isolation has probably been a dominant factor in the evolution of phylogeographically distinct groups on the North American coast, as well as in proximal terrestrial and freshwater systems (see Cox & Hebert 2001). Although little information on the temporal concordance of these patterns is presented here (see Cunningham & Collins 1994 for a review of the problem of pseudocongruence in phylogeography), it is most parsimonious given these data to assume that genetic drift is most responsible for the divergent populations seen. Clearly, there are differences in expected phylogeographical patterns given primary adaptive divergence and secondary contact between historical isolates (Endler 1977) but only further integration of research from a number of distinct programs will help to resolve this issue.

The strongest argument in favour of a historical influence on Northwest Atlantic phylogeography is the concordance of so many species. It is no longer satisfying to observe phylogeographical discontinuities in a single species, because only comparisons with other loci or other species can confirm that extrinsic forces have produced the discontinuity (Endler 1977; Hoelzer 1997; Avise 2000). Comparative phylogeography has also been useful in illuminating the history of geographical regions such as the Gulf Coast of North America (e.g. Avise 1992; O’Foighil et al. 1996; Herke & Foltz 2002) or the Pleistocene history of entire taxonomic groups (e.g. birds, Klicka & Zink 1997; Zink 1996; Holder et al. 1999; Edwards & Beerli 2000). The idea for comparing these discontinuities spatially is not new (see Nelson & Platnick 1981) but if we do it correctly we may shed light on the historical supply side dynamics of community assembly and evolution.

There is a true puzzle to be found in the boundaries between biogeographical provinces. That species with varied phylogenetic backgrounds, physiological tolerances and life history traits would tend to be co-distributed has begun to illuminate the importance of historical events on patterns of distribution and abundance, although it is clear that many more forces may be at work. This problem is particularly important because it connects the otherwise disparate fields of historical ecology and community assembly; the study of community genetics is one way of looking at this overlap to understand how ecological interactions and historical events have combined to create the patterns of diversity we see today, and to understand how these communities will respond to environmental change. Only by integrating information from a number of fields, including ecology, geology, oceanography and phylogeography, can we have a complete picture of these dynamic systems.

Acknowledgements

Many thanks to C. W. Cunningham, for keeping me afloat in these waters; M. J. Hickerson, C. Henzler, M. E. Hellberg, T. F. Turner, C. S. Embach and two anonymous reviewers all helped to dramatically improve the manuscript. J.P.W. was supported by a grant from the U.S.D.A. Forest Service.

John Wares received his PhD at Duke University after working on the reconstruction of postglacial intertidal communities in the North Atlantic using phylogeographical methods. This review is an exploration of some issues that were missing from his earlier publications, and fits into his postdoctoral research both at the University of New Mexico and (as of summer 2002) the University of California at Davis on the interactions of ecological dynamics, oceanographic or hydrological effects, and the molecular description of these dynamics.

Information Box Generating genetic diversity takes time

One way to determine the processes involved in phylogeographical divergence is to correlate genetic divergence with a temporal framework of environmental events (e.g. glacial maxima). Reliable estimates of substitution rates are not available for all taxa or genetic loci, so inferences must be made by considering the processes that actually generate genetic diversity and the time scale required for each to operate. On a short time scale (e.g. since the most recent glacial maximum) there is little reason to expect significant genetic divergence among populations. If we first choose to ignore the effects of mutation, the probability of fixing alternate alleles in different populations is low. Simulation data (below) show that, even for low effective population sizes (Ne), there is a high probability of two populations sharing an allele regardless of when the populations were separated. In the absence of mutation, the probability that the same allele is fixed in both populations is an asymptotic function over time inline imagefor a set of

populations with n initial alleles, each with a frequency pi. These data were generated using Ewens (1979) sampling formulae for the expected number and stationary frequency of alleles in an equilibrium population of given θ = 4Neμ, with the mutation rate μ = 1 × 10−5 substitutions/gene/generation. Allele frequencies are initially identical in each population; Monte Carlo resampling independently generates the allele frequency for each subsequent generation in each population. The time at which one or both populations is fixed for an allele no longer present in the other populations (due to drift, in the absence of mutation) is indicated for each of 100 replicates.

These plots illustrate previous theoretical work on the mean number of generations until an allele is lost when mutation is ignored. Crow & Kimura (1970; eqn 8.9.3) showed that this time t0(p) = −4Ne(p/(1 − p)) ln(p), where p is the initial frequency of the allele. Larger, more diverse populations may lose some alleles rapidly, but the time to fixation of an allele not found in a sister population is still great. Therefore, genetic drift alone will not generate significant divergence on short time scales (Hewitt 2000).Fig. 2

Figure 2.

Considering mutation, we can ask what historical pathway has allowed for divergence of two populations. In a contiguous range expansion such as that documented for many species following Pleistocene glaciation, the probability that an allele from the ‘source’ population still exists in an expansion population is high over time periods less than ≈ 50 000 years, assuming typical per-gene substitution rates (Maruyama & Fuerst 1984). A detailed model of this process was employed by Palumbi & Kessing (1991) to calculate the likelihood of observing identical sequences in two isolated populations of the sea urchin Strongylocentrotus over time. Their method accounts for rate variation among codon sites to calculate the probability of DNA sequence divergence over time, and shows that the probability of observing identical DNA sequences (for Strongylocentrotus, though these results are fairly general) is high for separation times up to ≈ 50 000 years. Beyond that time, there is a high likelihood of seeing genetic divergence between the populations. Therefore, populations that survived in glacial refugia should be more divergent from a

given population than those formed by a range expansion from the same index population, even though both will have reduced allelic diversity.

The most important factor in determining the age and cause of a genealogical pattern is the relationship of both drift and mutation to Ne. Edwards & Beerli (2000) use Bayesian methods to illustrate the importance of considering realistic estimates of Ne in estimating divergence times; a pattern that appears to be temporally concordant across a number of species should be closely examined for appropriate estimates of µ and Ne. Estimation of these parameters becomes more problematic as they are estimated simultaneously (Wakeley 2000), a statistical problem that may affect estimates of population age under varying patterns of migration and/or Ne. A simple method for estimating population divergence relies on the assumption that current Ne is an adequate representation of the ancestral Ne for a species. Nei & Li (1979) estimate of the net nucleotide divergence between two populations remains a reasonable, unbiased method for estimating the lower bound of divergence times, while accounting for the polymorphism in ancestral populations (reviewed in Edwards & Beerli 2000).

Ancillary